Panda Coder

YouTuber, Programmer, Sharing my journey through programming, startups, AI & more. Tg channel: pandaprogercore

[Panda Coder @PandaCoderexe]

Curious to see what awaits us tomorrow

I think, as always, a huge number of startup ideas and actual startups will die

My guess is that it’ll be something like the Grok bots, but from OpenAI.

What do you guys think?

[OpenAI @OpenAI]

Get ready.

[Panda Coder @PandaCoderexe]

We stayed up all night and did our best. And this is what came out of it

Huge shout-out to the whole team for making this happen

Here’s the result:

[iroh.hl (🥅,🥅) @0x_iroh]

We built a 3D tea shop with three.js for the Nansen Meridian Buildathon 🍵 Hope you guys enjoy it!

What we have live now:
• Read Uncle's 3 crypto theses scored by smart money
• Chat with Uncle, an AI host (Nansen Research agent under the hood)
• Meet the month's top HL traders, smart money, whales and memecoin traders.

This has been so fun and exciting to build!
Many, many AI tokens were used, a lot of polishing and testing but the result ended up being so nice, interactive and hopefully useful.

Team: @tldde @0x_Takezo @david_grii @PandaCoderexe

Built with @nansen_ai

Try it live: https://t.co/fjz0T6TASi
Github: https://t.co/cKu8vNZJAX

[Panda Coder @PandaCoderexe]

@br_huni Have you tested it on a real website?

[Panda Coder @PandaCoderexe]

Tested Sol 6 on the Plus plan, and the limit hits faster than I can blink.

Crazy 😭

[Panda Coder @PandaCoderexe]

We’ve reached the point where the OpenAI vs Anthropic race is causing way too many unnecessary arguments.

It starts to feel more like Pepsi vs Coke or Ronaldo vs Messi.

“Claude is better.”
“Codex is better.”

Like bro, chill out.

First of all, in a couple of months both companies will probably release even newer models, and they’ll still be pretty close to each other anyway.

Use whatever works best for you and stop trying to convince everyone else to use the same model just because you use it, or because it scored 1–2% higher on some benchmark.

[Claude @claudeai]

Introducing Claude Opus 5.5, the first model in our new Claude 5.5 family.

It performs at the level of Claude Fable 5.1 for most tasks, and costs 40% less to run than Opus 5.

[Panda Coder @PandaCoderexe]

Two frontier model drops in one day is already interesting, but the changes might be even more interesting.

Today OpenAI released GPT-6 Sol + Luna, and Anthropic released Claude Opus 5.5.
(Honorable mention to Grok 4.7, which launched yesterday.)
For OpenAI, what caught my attention the most is the price difference, especially considering that the models also improved in capability.

GPT-5.6 Sol -> GPT-6 Sol
$4/$20 -> $2/$10 per 1M input/output tokens

GPT-5.6 Luna -> GPT-6 Luna
$0.20/$1.20 -> $0.10/$0.50

At the same time, OpenAI reports better performance in coding, agentic tasks and factuality.

Anthropic also made a pretty big jump with Opus 5.5. It is around 20% cheaper per token, around 40% cheaper on typical workloads, and over 30% faster in output generation compared to Opus 5.

The benchmark improvements are also noticeable:

Terminal-Bench
Opus 5: 52.3%
Opus 5.5: 66.4%

FrontierCode
Opus 5: 48.0%
Opus 5.5: 54.4%

So right now we are getting more capable frontier models while the cost of actually using them is going down pretty fast.

I’m testing GPT-6 Sol right now and so far I think it fits the Plus plan really well, especially in terms of speed and general usability.

Can’t say much about intelligence yet though because I already hit the limits. I’ll have a much better idea tomorrow after I can test it properly.

Pretty interesting day for AI.

[ChatGPT @ChatGPT]

A new star enters the Chat.

Meet GPT-6 Astra.

[Panda Coder @PandaCoderexe]

It seems like Jev wasn’t even fully born before it already started dying. Just 4 days after Jev’s introduction, a Chinese team introduced its own self-hosted ML model that can be nearly 70× faster.

Of course, that speed comes at a cost, around 57% accuracy compared with 78% for Jev in one benchmark, but privacy and no API cost can still make Laya a very attractive option. Especially considering that, for example, Laya performs better in typed business and agent-decision benchmarks.

So which one would you choose?

[Panda Coder @PandaCoderexe]

Link to the interview: https://t.co/ntqc6JhKC0

[Panda Coder @PandaCoderexe]

After the recent news, there is really only one question: Do AI companies actually control their models, or are we slowly reaching a point where these systems can behave in ways their creators did not fully anticipate?

As ex-OpenAI researcher Daniel Kokotajlo says in his interview, AI companies do not pay sufficient attention to model safety and are just training and introducing new models as fast as possible to stay ahead in the market, without proper testing. He warns that if this continues, advanced AI systems could eventually pose an existential risk to humanity, including, in the worst-case scenario, human extinction.

Link to the full interview below.

Yesterday, Google confirmed that Gemini autonomously broke into the systems of three real companies during cybersecurity testing.

In one case, the model guessed passwords until it gained access. In two others, it found exposed credentials in public repositories and used them to enter protected systems.

This and other situations just confirm the fact that AI can be the largest digital weapon ever created if it falls into the wrong hands.

On the other hand, AI is one of the most impressive and greatest human inventions in history, which, in combination with other inventions like quantum computers, can help people, improve civilization, and basically make our lives better.

We are lucky to have the chance to live through and watch the progress of this whole thing in real time.

I’m keeping my fingers crossed that the second option plays out in the future and that people live in harmony and peace alongside AI robots.

Stay Tuned💻

[Panda Coder @PandaCoderexe]

Cooking with my colleagues: Codex and Claude Code

MVP coming online soon…

Stay tuned 💻

[Panda Coder @PandaCoderexe]

Salimpoor et al. (2011) Nature Neuroscience
Zatorre & Salimpoor (2013) PNAS
Salimpoor et al. (2009) PLOS ONE

[Panda Coder @PandaCoderexe]

Why do so many programmers wear headphones while coding?

Neuroscience has an interesting explanation.

I’ve noticed that almost every developer I know listens to music while programming. Personally, I usually code with handpan music.

So I started digging into the science.

[Panda Coder @PandaCoderexe]

In summary, music can improve mood, increase motivation, reduce stress and help you stay focused for longer.

Maybe that’s one of the reasons headphones have become almost part of every programmer’s setup.

What do you usually listen to while coding?